K. Resnicow et al., HOW BEST TO MEASURE IMPLEMENTATION OF SCHOOL-HEALTH CURRICULA - A COMPARISON OF 3 MEASURES, Health education research, 13(2), 1998, pp. 239-250
Citations number
53
Categorie Soggetti
Public, Environmental & Occupation Heath","Education & Educational Research
The impact of school health education programs is often attenuated by
inadequate teacher implementation. Using data from a school-based nutr
ition education program delivered in a sample of fifth graders, this s
tudy examines the discriminant and predictive validity of three measur
es of curriculum implementation: classroom observation of fidelity, an
d two measures of completeness, teacher self-report questionnaire and
post-implementation interview A fourth measure, obtained during teache
r observations, that assessed student and teacher interaction and stud
ent receptivity to the curriculum (labeled Rapport) was also obtained.
Predictive validity was determined by examining the association of im
plementation measures with three study outcomes; health knowledge, ask
ing behaviors related to fruit and vegetables, and fruit and vegetable
intake, assessed by 7-day diary. Of the 37 teachers observed, 21 were
observed for two sessions and 16 were observed once. Implementation m
easures were moderately correlated, an indication of discriminant vali
dity. Predictive validity analyses indicated that the observed fidelit
y, Rapport and interview measures were significantly correlated with p
ost-test student knowledge, The association between health knowledge a
nd observed fidelity (based on dual observation only), Rapport and int
erview measures remained significant after adjustment for pre-test kno
wledge values, None of the implementation variables were significantly
associated with student fruit and vegetable intake or asking behavior
s controlling for pre-test values. These results indicate that the tea
cher self-report questionnaire was not a valid measure of implementati
on completeness in this study. Post-implementation completeness interv
iews and dual observations of fidelity and Rapport appear to be more v
alid, and largely independent methods of implementation assessment.